New turn in US and China AI dialogue as Bessent names its aims

In September, during the visit of Xi, Washington and Beijing agreed to start an AI dialogue with a channel for serious incidents. That month the foreign ministry of China dismissed calls to slow down advanced models.

US Treasury Secretary Scott Bessent spoke in New York on Tuesday. He said that the US and China set up a new dialogue on “super intelligence”, his term for very advanced AI. The aim is to guard against shared risks and to prevent misunderstanding. He made no commitment to limit the technology, and both countries stay in a race for AI leadership.

What is new

• Both sides exchange best practices for testing advanced models before release.
• Both keep human judgment in command of the technology.
• Bessent expects both sides to alert each other to serious incidents.

The limits

No test results apply, as this is a statement of intent. Both countries remain in a race for AI leadership. The US restricts exports of the most advanced chips to China. President Trump said he is open to guardrails. He does not want the US in “a deep second position” behind China.

What is not known

No dates or rules for the dialogue are given. The two leaders are expected to meet on 18 and 19 November at the Asia-Pacific Economic Cooperation meeting. Bessent will join.

Insight: Both nations talk safety while they guard their own AI

The US will share testing practices, but its own testing unit CAISSI has $10 million a year and no permanent leader. China drafted export rules for 23 AI technologies, with licenses for model weights and agents. The dialogue seems a channel to manage incidents between rivals, not a brake on the race. Watch the meeting next week for a written rule.